Grapha AI

As of 31-Jan-2023, Grapha has acquired ~2k users and ~20 Grapha Prime (paid) users.

I spent the last few weeks improving data pre-processing and auto-cleaning bad datasets and improving error handling.

With Grapha's user base growing, I had to upgrade to a paid plan with Loops for emails. With users growing that also meant more time looking into support emails by investigating and responding to early adopters. This isn't a large problem at this point.

I also shipped Excel & Google Sheets support an made database optimizations to reduce my Bubble DB cost by moving datasets to Grapha's dedicated S3 bucket.

I also improved better logging and analytics in LogSnag.

Going to be spending the next few nights & weekends on marketing assets, copy & email campaigns.

Built and launched v1 of Grapha. Spent a few weeks getting feedback from friends & family and some paid reviews. Worked on refining the designs and branding.

I continued to do this on nights and weekend, making slow progress. The worst was when I went days without contributing to Grapha and it took longer catching up from where I'd left off and keeping track of bugs.

I decided if I were to ever ship Grapha, it would have to be during the holidays when I'd be able to dedicate extended hours and coding towards realistic milestones.

I got the product to a ready-to-ship state on 27-Dec-2023 and listed on https://theresanaiforthat.com/ai/grapha-ai/

This led to Grapha getting it's first user and first early bird subscription within the first day.

It didn't take long before I realized once again how much of a real problem it was for me to spend hours after work catching up on those data deep dives.

At this point, the frustrations that led to killing the app a month ago wore off and I was willing to give Grapha the nurturing it needed to get to to market in a month.

It didn't just take a month.

I started building the product and the frontend experience. I have zero experience building frontend with code. For all my web apps, I've always resorted to Bubble.io and while it is great and I would recommend investing the time to build with bubble to any starting founder, I quickly realized when building Grapha the limitations of my lack of experience building frontend experiences would have on building and scaling Grapha. Or at least I thought at the time.

I drop the idea entirely and shut down my bubble project.

I built a very stripped down MVP in Jupyter notebook that accepted light CSVs (Kaggle probably thinks I'm a bot looking for flat datasets); and automatically generated insights with some guidance parameters from the user.

This worked extremely well and my OpenAI API cost was under what I was expecting. Spent the next few nights & weekends learning about prompt chaining.

Things were looking promising.

I've had experience building products - the most recent at the time being https://getunfold.app (shutdown) & https://bravokit.app; which allows builders and product teams to host their product roadmap online to welcome users to submit feature requests and vote on what the project was considering to build next.

Like I mentioned, I also had a data science background and knew what needed to be built as an MVP. I spent quite a few weeks brushing up on Python, pandas & FastAPI while wireframing the frontend experience & backend workflows.

Although I didn't have a clear concept of the product at the time, in my product management role at my day job, I found myself being inundated with a lot of data that needed analysis.

While I came from a data background and was technically equipped to do the analysis myself, I simply didn't have the time to do them quick enough. Raising a ticket to BI often meant scheduling a call to explain my ask, all of which took as much time as running the analysis myself.